Machine learning of noise-resilient quantum circuits

In this work, we study how machine learning can be applied to formulate noise-aware circuit compilations that can be executed on near-term quantum hardware to produce reliable results. We will demonstrate that experimentally derived noise models can be used to go beyond naive circuit compilations for several example quantum algorithms. There are two inputs to our Noise-Aware Circuit Learning (NACL) method: a task, and a noisy gate alphabet. The task is defined by either a set of classical training data or a desired output quantum state or unitary. The output of NACL is a quantum gate sequence that optimally accomplishes the inputted task in the presence of the inputted noise model. Neither an ansatz nor the circuit depth of the gate sequence is an input to NACL. This is because NACL optimizes over the circuit structure and depth, which is in the spirit of task-oriented programming. We implement NACL for several different problems, such as computing state overlap, preparing multi-body entangled states, and implementing the quantum Fourier transform. In each case, we find that our overall figure-of-merit is significantly lower for NACL than for standard methods of circuit compilation.